Silty clay layer soft foundation treatment method based on extension rod mixing pile
The method of connecting drill rods for mixing piles, which involves real-time acquisition of soil parameters and the use of LSTM models for risk prediction and adaptive control, solves the problems of large dispersion in pile strength and long connection time caused by abrupt changes in soil layers in traditional mixing pile construction. This method achieves stable connection of drill rods and improves construction efficiency.
Patent Information
- Application Number
- CN202511011861.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-22
- Publication Date
- 2025-12-02
AI Technical Summary
Traditional mixing pile construction cannot adapt to sudden changes in soil layers, resulting in large dispersion in pile strength. The splicing process requires manual operation after machine shutdown, which increases downtime and is prone to pile breakage and grout segregation.
Soil parameters are collected in real time using rheological property sensors, micro shear strength sensors, and microwave moisture sensors. Risk prediction and adaptive control are performed using an LSTM model based on edge computing. Stable docking of drill rods is achieved by combining the drill rod docking device.
It enables real-time response to sudden changes in soil layers, reduces manual operation time, lowers the probability of drill rod jamming, and improves the stability of pile strength and construction efficiency.
Smart Images
Figure CN121047255A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of mixing pile construction technology, and in particular to a method for treating soft soil foundations in silty clay layers based on spliced mixing piles. Background Technology
[0002] Mixing pile construction is a foundation treatment technology. Its core is to use deep mixing method to forcibly mix the solidifying agent (usually cement slurry or lime powder) with the soft soil (such as silt, silty soil, cohesive soil, silty soil, loose sand, etc.) in the foundation in situ.
[0003] Traditional mixing pile construction parameters (rotation speed, grouting pressure, etc.) rely on experience to be set, which cannot adapt to sudden changes in soil layers (such as the alternation of silty clay and sand layers), resulting in large dispersion of pile strength. The splicing process requires manual operation and shutdown, and manual fixation of the splicing drill rod is required. The splicing time is relatively long, increasing downtime and easily causing pile breakage and grout segregation. Existing monitoring technologies (such as penetration tests) are lagging behind and cannot provide real-time feedback on soil layer characteristics and pile quality. Summary of the Invention
[0004] The existing mixing pile construction methods cannot adapt to abrupt changes in soil layers (such as silty clay and sand layers), resulting in large dispersion of pile strength. The splicing process requires manual operation and shutdown, and manual fixing of the splicing drill rods, which takes a long time and increases downtime. This can easily lead to technical problems such as pile breakage and grout segregation. Therefore, this invention proposes a method for treating soft soil foundations in silty clay layers based on splicing mixing piles.
[0005] The present invention proposes a method for treating soft soil foundation of silty clay layer based on joint rod mixing pile, which includes the following steps: Step 1: Start the joint rod mixing pile machine to drive the drill rod to drill downward. While the drill rod is drilling, soil parameters are collected by rheological property sensor installed on the drill rod and miniature shear strength sensor and microwave moisture content sensor on the drill bit. The collected data is uploaded to the edge computing gateway in real time through wireless transmission module.
[0006] Step 2: Preprocess and extract temporal features from the collected data: Perform sliding window denoising and feature engineering denoising on the raw data.
[0007] Step 3: Real-time inference of the LSTM risk prediction model: The edge computing gateway loads the pre-trained LSTM model.
[0008] Step 4: Risk Classification and Adaptive Control.
[0009] Step 5: Online updating and feedback optimization of the established model: After every 10 drilling cycles, the field data (sensor data + actual construction parameters) is synchronized to the cloud training platform, and incremental learning is used to update the LSTM model weights and optimize the mapping relationship.
[0010] Preferably, in step one, the data collected by the micro shear strength sensor is used to measure the shear strength of the soil at the drill bit contact surface, the data collected by the microwave moisture content sensor is used to monitor the real-time moisture content of the soil, and the data collected by the rheological property sensor is used to record the creep rate of the soil.
[0011] Preferably, the sliding window denoising formula in step two is: Where χ t This is the current sampled value. The filtered value is α, which is a smoothing factor and is set to 0.2.
[0012] Extracting temporal features includes the rate of change of shear intensity within 5 seconds. Δτ rate of change of shear strength The instantaneous rate of change of shear strength, and the covariance of moisture content and creep rate.
[0013] 10s moving standard deviation of rheological rate Let i be the rheological rate at the i-th sampling point. This represents the average rheological rate.
[0014] Preferably, the step of loading the pre-trained LSTM model in step three includes: Input layer: receiving the pre-processed feature vectors. τ represents the real-time shear strength, measured directly by a shear strength sensor, and W represents the real-time moisture content. Real-time rheological rate;
[0015] Hidden layer: Learns temporal dependencies through two layers of LSTM units, and outputs the hidden state h. t ;
[0016] Output layer: Calculate the stress concentration risk index η, the calculation formula is: η=σ(W o ·h t +b o ), η∈[0,1], where σ is the activation function, and the function formula is: e is the base of the natural logarithm, z = W O ·h t+ b o W O h is the output layer weight matrix. t For the hidden state at time step t, the feature vector output by the LSTM memory unit encodes the temporal features of the sensor, b. o The parameters for adjusting the risk baseline are used to set the output layer bias vector.
[0017] Preferably, in step four, a graded response is triggered based on η, where η < 0.5 indicates low risk, and the current construction parameters are maintained.
[0018] For a medium risk level of 0.5 ≤ η < 0.7, reduce torque by 10% and increase grouting pressure by 5%.
[0019] If η ≥ 0.7, it indicates high risk. The following actions should be taken: First, avoid the pole connection point: calculate the safe connection point using a path planning algorithm to avoid stress concentration areas; second, adaptively adjust the torque according to formula T. new =T0·(1-0.3η) Dynamic torque reduction, T new The adjusted torque is T0, which is the reference torque. Finally, the amount of curing agent is increased by 15% to enhance the stability of the pore walls.
[0020] Preferably, the formula for optimizing the mapping relationship in step five is: θ new The weights refer to the old model weights, specifically the current parameter matrix of the LSTM neural network, which includes all weights and biases from the input layer to the hidden layer to the output layer, θ. old Here, λ represents the weights of the new model, specifically referring to the network parameters updated through incremental learning. λ is the learning rate, a hyperparameter that controls the step size of parameter updates; a smaller value indicates a more conservative update. The gradient operator specifically refers to calculating the partial derivative of the loss function with respect to the weights θ, where L is the loss function, calculated using the following formula: n is the sample size, Y pred For the model's prediction output, the LSTM predicts the following construction parameter vectors: rotation speed, grouting pressure, hardener dosage, rotation speed, grouting pressure, hardener dosage, Y. true These are the actual construction parameters, the parameters that were actually executed (recorded by the PLC control system).
[0021] Preferably, in step one, a connecting rod device is provided on the outer surface of the support of the connecting rod mixing pile machine. The connecting rod device includes a left wedge and a right wedge. The inner wall of the drill rod has a docking hole. One left wedge and one right wedge are slidably inserted into the inner wall of the docking hole. After the two sections of one left wedge and one right wedge are docked, a pin is slidably inserted into the inner wall. The pin is slidably inserted into the inner wall of the drill rod. The outer surface of the pin has a fixing groove. The inner walls of the left wedge and the right wedge are threaded with fixing bolts. One end of the fixing bolt is rotatably connected to a fixing arc block. One end of the two fixing arc blocks is slidably inserted into the inner wall of one left wedge and one right wedge, respectively. The inner surface of the fixing arc block is slidably inserted into the inner wall of the fixing groove.
[0022] Preferably, the connecting rod device further includes a support ring, which is fixedly installed on the outer surface of the support frame of the connecting rod mixing pile machine. A positioning hydraulic cylinder arranged in a ring array is fixedly installed on the outer surface of the support ring. A positioning block with a ball bearing is fixedly installed at one end of the piston rod of the positioning hydraulic cylinder. A distance sensor is fixedly installed on the inner surface of the support ring.
[0023] Preferably, a symmetrically distributed holding groove is fixedly installed on the outer surface of the support ring, a pushing hydraulic cylinder with a push plate is fixedly installed on the lower surface of the holding groove, a moving component is fixedly installed on the outer side of the holding groove, a rotary motor is fixedly installed on the outer surface of the slider of the moving component, a push rod is fixedly installed at one end of the output shaft of the rotary motor, and one end of the push rod is slidably inserted into one end of the fixing bolt.
[0024] Preferably, a cooling housing is fixedly installed on the outer surface of the support ring, a lifting hydraulic cylinder with a push plate is fixedly installed on the lower surface of the cooling housing, a cover plate is hinged to the outer surface of the cooling housing by a pin, a cooling guide plate is fixedly installed on the inner wall of the cooling housing, cooling fins are fixedly installed inside the cooling housing, a pushing hydraulic cylinder is fixedly installed on the outer surface of the cooling housing, a support sleeve is slidably inserted into the inner wall of the support ring, a semi-rotating ring with an arc groove is rotatably connected to the upper surface of the support ring, the inner wall of the semi-rotating ring is slidably inserted into the upper surface of the support sleeve by a column, and the inner wall of the arc groove of the support ring is slidably inserted into one end of the piston rod of the positioning hydraulic cylinder by a column.
[0025] The beneficial effects of this invention are as follows:
[0026] 1. By setting processing methods, multi-source data fusion can be performed. By leveraging the coupling characteristics of shear strength mutation and creep rate increase, stress concentration can be predicted 5-8 seconds in advance. Secondly, real-time edge decision-making can be performed with an inference delay of <50ms, meeting the real-time control requirements of the drilling process. Finally, adaptive avoidance and drill rod path planning based on risk index can reduce the probability of drill rod jamming. This solves the technical problem that existing mixing pile machines cannot adapt to soil layer mutations (such as the intersection of silty clay and sand layers), resulting in large dispersion of pile strength.
[0027] 2. By setting up a rod-connecting device, manual labor can be reduced, thereby speeding up the rod-connecting process. The connecting rod is moved directly by a push rod, which pushes the right and left wedge blocks to be inserted into the drill rod through the docking holes. A pin is inserted into the through hole at the docking end of the right and left wedge blocks to fix the left and right wedge blocks and the drill rod. The pin is cooled and shrunk to facilitate insertion into the rotating rod, left and right wedge blocks. After the fixing bolt is rotated, the fixing arc block can be pushed into the fixing groove of the pin to fix the pin. This multiple fixing method maintains the stability of the drill rod connection and solves the technical problems of existing mixing pile machines requiring machine stoppage and manual operation during rod connection, which increases downtime and is prone to pile breakage and slurry segregation. Attached Figure Description
[0028] Figure 1 This is a schematic diagram of a method for treating soft soil foundations in silty clay layers based on a joint-type mixing pile, as proposed in this invention.
[0029] Figure 2 This is a three-dimensional view of the rheological property sensor structure for a soft soil foundation treatment method based on a joint-type mixing pile proposed in this invention.
[0030] Figure 3 This is a perspective view of the left wedge structure of a soft soil foundation treatment method for silty clay layers based on a joint-pile mixing pile proposed in this invention.
[0031] Figure 4 This is a perspective view of the pin structure of a soft soil foundation treatment method for silty clay layer based on a joint-type mixing pile proposed in this invention.
[0032] Figure 5 This is a perspective view of the fixing bolt structure of a soft soil foundation treatment method for silty clay layer based on a joint-type mixing pile proposed in this invention.
[0033] Figure 6 This is a perspective view of the positioning hydraulic cylinder structure of a soft soil foundation treatment method for silty clay layers based on a joint-type mixing pile proposed in this invention.
[0034] Figure 7 This is a perspective view of the push rod structure of a soft soil foundation treatment method for silty clay layer based on a joint-mixing pile proposed in this invention.
[0035] Figure 8 This is a three-dimensional view of the semi-circular structure of a soft soil foundation treatment method for silty clay layers based on a joint-type mixing pile proposed in this invention.
[0036] Figure 9 This is a perspective view of the support sleeve structure for a soft soil foundation treatment method based on a joint-pile mixing pile proposed in this invention.
[0037] Figure 10 This is a perspective view of the cooling plate structure of a soft soil foundation treatment method for silty clay layers based on a joint-pile mixing pile proposed in this invention.
[0038] In the diagram: 1. Jointed mixing pile machine; 11. Drill rod; 12. Rheological property sensor; 13. Miniature shear strength sensor; 14. Microwave moisture content sensor; 2. Left wedge; 21. Right wedge; 22. Docking hole; 23. Pin; 24. Fixing groove; 25. Fixing bolt; 26. Fixing arc block; 3. Support ring; 31. Positioning hydraulic cylinder; 32. Positioning block; 33. Distance sensor; 4. Container trough; 41. Pushing hydraulic cylinder; 42. Moving component; 43. Rotary motor; 44. Push rod; 5. Refrigeration shell; 51. Lifting hydraulic cylinder; 52. Cover plate; 53. Cooling plate; 54. Refrigeration element; 55. Pushing hydraulic cylinder; 56. Support sleeve; 57. Semi-rotating ring. Detailed Implementation
[0039] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.
[0040] Reference Figures 1-10 A method for treating soft soil foundations in silty clay layers based on a joint-type mixing pile includes the following steps: Step 1: Start the joint-type mixing pile machine 1 to drive the drill rod 11 to drill downwards. While the drill rod 11 is drilling, soil parameters are collected by a rheological property sensor 12 installed on the side wall of the drill rod 11 at a distance of 1.5M from the drill bit, as well as a miniature shear strength sensor 13 and a microwave moisture content sensor 14 on the drill bit of the drill rod 11. The collected data is uploaded to the edge computing gateway in real time through a wireless transmission module.
[0041] In step one, the data collected by the miniature shear strength sensor 13 is used to measure the shear strength of the soil at the drill bit contact surface, the data collected by the microwave moisture content sensor 14 is used to monitor the real-time moisture content of the soil, and the data collected by the rheological property sensor 12 is used to record the creep rate of the soil.
[0042] Step 2: Preprocess and extract temporal features from the collected data: Perform sliding window denoising and feature engineering denoising on the raw data.
[0043] The sliding window denoising formula in step two is: Where χ t This is the current sampled value. The filtered value is α, which is a smoothing factor and is set to 0.2.
[0044] Extracting temporal features includes the rate of change of shear intensity within 5 seconds. Δτ rate of change of shear strength The instantaneous rate of change of shear strength, and the covariance of moisture content and creep rate.
[0045] 10s moving standard deviation of rheological rate Where N = 10, and N is the number of data points used to calculate the standard deviation. Let i be the rheological rate at the i-th sampling point. This represents the average rheological rate.
[0046] Δτ and The core early warning indicator is: when Δτ < -20 kPa / s and When the probability of triggering a high risk is >90%, N=10 (5s window): response time of creep characteristics of the overburden layer.
[0047] Step 3: Real-time inference of the LSTM risk prediction model: The edge computing gateway loads the pre-trained LSTM model.
[0048] Step 3 involves loading the pre-trained LSTM model: Input layer: receives the pre-processed feature vectors. τ represents the real-time shear strength, measured directly by a shear strength sensor, and W represents the real-time moisture content. Real-time rheological rate (creep rate), Δτ rate of change of shear strength;
[0049] Hidden layer: Learns temporal dependencies through two layers of LSTM units, and outputs the hidden state h. t Output layer: Calculate the stress concentration risk index η, the calculation formula is: η=σ(W o ·h t +b o ), η∈[0,1](>0.7 is considered high risk), where σ is the activation function, and the function formula is: e is the base of the natural logarithm, z = W O ·h t+ b o W O h is the output layer weight matrix. t For the hidden state at time step t, the feature vector output by the LSTM memory unit encodes the temporal features of the sensor, b. o The parameters for adjusting the risk baseline are used to set the output layer bias vector.
[0050] Step 4: Risk Classification and Adaptive Control.
[0051] In step four, a graded response is triggered based on η. If η < 0.5, it is considered low risk, and the current construction parameters are maintained.
[0052] For a medium risk level of 0.5 ≤ η < 0.7, reduce torque by 10% and increase grouting pressure by 5%.
[0053] If η ≥ 0.7, it indicates high risk. The following actions should be taken: First, avoid the pole connection point: calculate the safe connection point using a path planning algorithm to avoid stress concentration areas; second, adaptively adjust the torque according to formula T. new =T0·(1-0.3η) Dynamic torque reduction, T new The adjusted torque is T0, which is the reference torque. Finally, the amount of curing agent is increased by 15% to enhance the stability of the pore walls.
[0054] Step 5: Online updating and feedback optimization of the established model: After every 10 drilling cycles, the field data (sensor data + actual construction parameters) is synchronized to the cloud training platform, and incremental learning is used to update the LSTM model weights and optimize the mapping relationship.
[0055] The formula for optimizing the mapping relationship in step five is: θ new The weights refer to the old model weights, specifically the current parameter matrix of the LSTM neural network, which includes all weights and biases from the input layer to the hidden layer to the output layer, θ. old Here, λ represents the weights of the new model, specifically referring to the network parameters updated through incremental learning. λ is the learning rate, a hyperparameter that controls the step size of parameter updates; a smaller value indicates a more conservative update. The gradient operator specifically refers to calculating the partial derivative of the loss function with respect to the weights θ, where L is the loss function, calculated using the following formula: n is the sample size, Y pred For the model's prediction output, the LSTM predicts the following construction parameter vectors: rotation speed, grouting pressure, hardener dosage, rotation speed, grouting pressure, hardener dosage, Y. true These are the actual construction parameters, the parameters that were actually executed (recorded by the PLC control system).
[0056] When encountering a water-rich, flowing sand layer (W>30%), the model automatically adjusts the grouting pressure weighting coefficient by +15% within 10 sets of data.
[0057] An incremental update is triggered after each drilling cycle (approximately 2 hours) is completed, while retaining 10% of the historical core samples to prevent drift.
[0058] By setting the processing method, it is possible to perform multi-source data fusion, predict stress concentration 5-8 seconds in advance by using the coupling characteristics of shear strength mutation and creep rate increase, secondly, it is possible to make edge real-time decisions with inference delay <50ms, meet the real-time control requirements of drilling process, and finally, it is possible to adaptively avoid obstacles and reduce the probability of drill rod (11) jamming based on the connection path planning of risk index. It solves the technical problem that the existing mixing pile machine construction cannot adapt to soil layer mutation (such as silty clay and sand layer), resulting in large dispersion of pile strength.
[0059] like Figures 1-5 As shown, to reduce the time for connecting drill rod 11, a connecting device is installed on the outer surface of the support of the connecting pile machine 1 in step one. The connecting device includes a left wedge 2 and a right wedge 21. A docking hole 22 is opened on the inner wall of the drill rod 11. One left wedge 2 and one right wedge 21 are slidably inserted into the inner wall of the docking hole 22. The left wedge 2 and the right wedge 21 can be inserted to form a complete wedge, which is then inserted into the docking hole 22 to fix the overlapping part of the two connecting rods. In order to maintain the stable fixation of the left wedge 2 and the right wedge 21, a pin 23 is slidably inserted into the inner wall of the two sections of the left wedge 2 and the right wedge 21 after docking. The material of the pin 23 is 40CrNiMoA, and the diameter is Φ20 at room temperature. -0.013 The interference fit is 0.013 to 0.025 mm. The pin 23 is slidably inserted into the inner wall of the drill rod 11. In order to stably fix the pin 23, a fixing groove 24 is provided on the outer surface of the pin 23. The inner walls of the left wedge 2 and the right wedge 21 are threaded with fixing bolts 25. One end of the fixing bolt 25 is rotatably connected to a fixing arc block 26. One end of the two fixing arc blocks 26 is slidably inserted into the inner wall of a left wedge 2 and a right wedge 21, respectively. The inner surface of the fixing arc block 26 is slidably inserted into the inner wall of the fixing groove 24.
[0060] like Figures 6-10 As shown, specifically, in order to fix the drill rod 11, the rod connecting device also includes a support ring 3. The support ring 3 is fixedly installed on the outer surface of the bracket of the rod connecting mixing pile machine 1. Positioning hydraulic cylinders 31 arranged in a ring array are fixedly installed on the outer surface of the support ring 3. A positioning block 32 with ball bearings is fixedly installed at one end of the piston rod of the positioning hydraulic cylinder 31. In order to accurately position the drill rod 11, a distance sensor 33 is fixedly installed on the inner surface of the support ring 3.
[0061] Specifically, in order to reduce manual labor input, a symmetrically distributed holding groove 4 is fixedly installed on the outer surface of the support ring 3. A pusher hydraulic cylinder 41 with a push plate is fixedly installed on the lower surface of the holding groove 4. A moving component 42 is fixedly installed on the outer side of the holding groove 4. The moving component 42 consists of a frame, a lead screw rotatably connected to the frame, a motor driving the lead screw to rotate, and a slider threadedly connected to the lead screw and slidably inserted into the frame. A rotary motor 43 is fixedly installed on the outer surface of the slider of the moving component 42. A push rod 44 is fixedly installed at one end of the output shaft of the rotary motor 43. One end of the push rod 44 is slidably inserted into one end of the fixing bolt 25.
[0062] Specifically, in order to cool down and shrink the pin 23, a cooling shell 5 is fixedly installed on the outer surface of the support ring 3. A lifting hydraulic cylinder 51 with a push plate is fixedly installed on the lower surface of the cooling shell 5. The push plate is fixedly installed on one end of the piston rod of the lifting hydraulic cylinder 51. A cover plate 52 is hinged to the outer surface of the cooling shell 5 by a pin. A cooling guide plate 53 is fixedly installed on the inner wall of the cooling shell 5. A cooling chip 54, which is a semiconductor cooling chip, is fixedly installed inside the cooling shell 5. A pushing hydraulic cylinder 55 is fixedly installed on the outer surface of the cooling shell 5. A support sleeve 56 is slidably inserted into the inner wall of the support ring 3. A semi-rotary ring 57 with an arc groove is rotatably connected to the upper surface of the support ring 3. The inner wall of the semi-rotary ring 57 is slidably inserted into the upper surface of the support sleeve 56 through a column. The inner wall of the arc groove of the support ring 3 is slidably inserted into one end of the piston rod of the positioning hydraulic cylinder 31 through a column.
[0063] Working principle: Soil parameters are collected by the variable characteristic sensor on the drill rod 11, the miniature shear strength sensor 13 on the drill bit, and the microwave moisture content sensor 14. The data is transmitted to the edge counting gateway, which prevents the drill rod from stopping in dangerous areas. When a drill rod needs to be connected, the positioning hydraulic cylinder 31 pushes the positioning block 32 to move. The positioning block 32 positions and clamps the drill rod 11 to be connected, keeping the center of gravity downwards, and inserts it with the drill rod 11 below. At the same time, the pushing of the positioning hydraulic cylinder 31 causes the semi-rotating ring 57 to rotate on the support ring 3, driving the support sleeve 56 to move outwards. After the left wedge 2 and right wedge 21 in the holding trough 4 are pushed up by the two pushing hydraulic cylinders 41, the moving component 42 is activated, driving the slider to move, so that the two pushing rods 44 are respectively inserted into the left wedge 2 and right wedge 21 and connected to the internal fixing bolts 25. The movement of the slider pushes the pushing rods 44, so that the left wedge 2 and right wedge 21 are connected to the support ring 3. After the right wedge 21 is inserted into the docking hole 22, it enters the drill rod 11 and interlocks with each other. The cooling plate 54 in the cooling shell 5 cools the pin 23, which is then cooled by the cooling plate 53. This reduces the temperature of the pin 23, causing it to shrink and position. The hydraulic cylinder 55 pushes the cooled pin 23 to move, opening the cover plate 52 and inserting it into the support sleeve 56. The pin 23 then passes through the through hole on the drill rod 11 and enters the left wedge 2 and right wedge 21, exiting from the other side of the drill rod 11. The pin 23 returns to its original shape at room temperature, expands, and is then press-fitted with the drill rod 11. At the same time, the rotating motor 43 drives the push rod 44 to rotate, which in turn drives the fixing bolt 25 to rotate. The screwing of the fixing bolt 25 presses against the fixed arc block 26, allowing the fixed arc block 26 to be inserted into the fixing groove 24 of the pin 23, thus providing multiple fixation for the pin 23 and maintaining the stability of the drill rod 11.
[0064] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.
Claims
1. A method for treating soft soil foundations in silty clay layers based on joint-pile mixing, characterized in that: The steps include: Step 1: Start the joint mixing pile machine (1) to drive the drill rod (11) to drill downwards. While the drill rod (11) is drilling, soil parameters are collected by the rheological property sensor (12) installed on the side wall of the drill rod (11) and the micro shear strength sensor (13) and microwave moisture content sensor (14) on the drill bit of the drill rod (11). The collected data is uploaded to the edge computing gateway in real time through the wireless transmission module. Step 2: Preprocessing and extracting temporal features from the collected data: Perform sliding window denoising and feature engineering denoising on the raw data; Step 3: Real-time inference of the LSTM risk prediction model: The edge computing gateway loads the pre-trained LSTM model; Step 4: Risk Classification and Adaptive Control; Step 5: Online update and feedback optimization of the established model: After every 10 drilling cycles, the field data is synchronized to the cloud training platform, and incremental learning is used to update the LSTM model weights and optimize the mapping relationship.
2. The method for treating soft soil foundations in silty clay layers based on jointed mixing piles according to claim 1, characterized in that: In step one, the data collected by the micro shear strength sensor (13) is used to measure the shear strength of the soil at the drill bit contact surface, the data collected by the microwave moisture content sensor (14) is used to monitor the real-time moisture content of the soil, and the data collected by the rheological property sensor (12) is used to record the creep rate of the soil.
3. The method for treating soft soil foundations in silty clay layers based on jointed mixing piles according to claim 2, characterized in that: The sliding window denoising formula in step two is as follows: Where χ t This is the current sampled value. The filtered value is α, which is a smoothing factor and is set to 0.
2. Extracting temporal features includes the rate of change of shear intensity within 5 seconds. Δτ rate of change of shear strength The instantaneous rate of change of shear strength, and the covariance of moisture content and creep rate. W represents the real-time moisture content. Real-time rheological rate; 10s moving standard deviation of rheological rate Where N = 10, and N is the number of data points used to calculate the standard deviation. Let i be the rheological rate at the i-th sampling point. This represents the average rheological rate.
4. The method for treating soft soil foundations in silty clay layers based on jointed mixing piles according to claim 3, characterized in that: The step of loading the pre-trained LSTM model in step three: Input layer: receives the pre-processed feature vectors. τ represents the real-time shear strength, which is directly measured by the shear strength sensor. Hidden layer: Learns temporal dependencies through two layers of LSTM units, and outputs the hidden state h. t ; Output layer: Calculate the stress concentration risk index η, the calculation formula is: η=σ(W o ·h t +b o ), η∈[0,1], where σ is the activation function, and the function formula is: e is the base of the natural logarithm, z = W O ·h t+ b o W O h is the output layer weight matrix. t For the hidden state at time step t, the feature vector output by the LSTM memory unit encodes the temporal features of the sensor, b. o The parameters for adjusting the risk baseline are used to set the output layer bias vector.
5. The method for treating soft soil foundations in silty clay layers based on jointed mixing piles according to claim 4, characterized in that: In step four, a graded response is triggered based on η. If η < 0.5, it indicates low risk, and the current construction parameters are maintained. For a medium risk level of 0.5 ≤ η < 0.7, reduce torque by 10% and increase grouting pressure by 5%. If η≥0.7 is considered high risk, the following actions should be taken: First, avoid the pole connection point: Calculate the safe pole connection point using a path planning algorithm to avoid stress concentration areas; Secondly, torque adaptive adjustment: according to formula T new =T0·(1-0.3η) Dynamic torque reduction, T new The adjusted torque is T0, which is the reference torque. Finally, the amount of curing agent is increased by 15% to enhance the stability of the pore walls.
6. A method for treating soft soil foundations in silty clay layers based on a joint-pile mixing system according to claim 5, characterized in that: The formula for optimizing the mapping relationship in step five is: θ new The weights refer to the old model weights, specifically the current parameter matrix of the LSTM neural network, which includes all weights and biases from the input layer to the hidden layer to the output layer, θ. old Here, λ represents the weights of the new model, specifically referring to the network parameters updated through incremental learning. λ is the learning rate, a hyperparameter that controls the step size of parameter updates; a smaller value indicates a more conservative update. The gradient operator specifically refers to calculating the partial derivative of the loss function with respect to the weights θ, where L is the loss function, and the calculation formula is: n is the sample size, Y pred For the model's prediction output, the LSTM predicts the following construction parameter vectors: rotation speed, grouting pressure, hardener dosage, rotation speed, grouting pressure, hardener dosage, Y. true These are the actual construction parameters, the parameters that were actually implemented.
7. The method for treating soft soil foundations in silty clay layers based on jointed mixing piles according to claim 1, characterized in that: In step one, a connecting rod device is provided on the outer surface of the support of the connecting rod mixing pile machine (1). The connecting rod device includes a left wedge (2) and a right wedge (21). A docking hole (22) is opened on the inner wall of the drill rod (11). One left wedge (2) and one right wedge (21) are slidably inserted into the inner wall of the docking hole (22). After the two sections of one left wedge (2) and one right wedge (21) are docked, a pin (23) is slidably inserted into the inner wall. The pin (23) is connected to the drill rod (11). The inner wall of the pin (23) is slidably inserted, and a fixing groove (24) is provided on the outer surface of the pin (23). The inner walls of the left wedge (2) and the right wedge (21) are threaded with fixing bolts (25). One end of the fixing bolt (25) is rotatably connected to a fixing arc block (26). One end of the two fixing arc blocks (26) is slidably inserted into the inner wall of one of the left wedges (2) and one of the right wedges (21), respectively. The inner surface of the fixing arc block (26) is slidably inserted into the inner wall of the fixing groove (24).
8. The method for treating soft soil foundations in silty clay layers based on joint-pile mixing as described in claim 7, characterized in that: The connecting rod device also includes a support ring (3), which is fixedly installed on the outer surface of the support frame of the connecting rod mixing pile machine (1). A positioning hydraulic cylinder (31) arranged in a ring array is fixedly installed on the outer surface of the support ring (3). A positioning block (32) with a ball bearing is fixedly installed at one end of the piston rod of the positioning hydraulic cylinder (31). A distance sensor (33) is fixedly installed on the inner surface of the support ring (3).
9. A method for treating soft soil foundations in silty clay layers based on a joint-pile mixing system according to claim 8, characterized in that: The outer surface of the support ring (3) is fixedly equipped with symmetrically distributed holding grooves (4). The lower surface of the holding groove (4) is fixedly equipped with a pusher hydraulic cylinder (41) with a pusher plate. The outer side of the holding groove (4) is fixedly equipped with a moving component (42). The outer surface of the slider of the moving component (42) is fixedly equipped with a rotary motor (43). One end of the output shaft of the rotary motor (43) is fixedly equipped with a push rod (44). One end of the push rod (44) is slidably inserted into one end of the fixing bolt (25).
10. A method for treating soft soil foundations in silty clay layers based on a jointed mixing pile according to claim 9, characterized in that: A cooling shell (5) is fixedly installed on the outer surface of the support ring (3). A lifting hydraulic cylinder (51) with a push plate is fixedly installed on the lower surface of the cooling shell (5). A cover plate (52) is hinged to the outer surface of the cooling shell (5) by a pin. A cooling plate (53) is fixedly installed on the inner wall of the cooling shell (5). A cooling chip (54) is fixedly installed inside the cooling shell (5). A pushing hydraulic cylinder (55) is fixedly installed on the outer surface of the cooling shell (5). A support sleeve (56) is slidably inserted into the inner wall of the support ring (3). A semi-rotating ring (57) with an arc groove is rotatably connected to the upper surface of the support ring (3). The inner wall of the semi-rotating ring (57) and the upper surface of the support sleeve (56) are slidably inserted through a column. The inner wall of the arc groove of the support ring (3) and one end of the piston rod of the positioning hydraulic cylinder (31) are slidably inserted through a column.